Building a SaaS in the SaaS Apocalypse
Earlier this year, SaaS valuations dropped across the board. The models got smart enough that people could build their own tools. Agent-driven development made it accessible. Founders paying hundreds a month across fragmented platforms started asking: why am I paying for this?
For a lot of tools, that's a fair question. A simple project tracker, a basic form builder — if a founder can spin one up for their specific needs, the pricing model gets hard to justify. Enterprise and collaborative use cases still need the cloud and the sophistication. But for a wide range of users, the floor dropped.
We launched a SaaS anyway. Here's the thinking.
How It Started
My co-founder Louie runs a design agency in the Netherlands. Six years, real clients, real team. He was paying hundreds of euros across ClickUp, Typeform, Slack, Google Drive, invoicing tools, review tools, time trackers. None built for how design agencies work.
Louie worked with Claude Code to produce an MVP in a few days. He got a shell — enough features to illustrate his vision. Nothing worked well yet, but the product thinking was clear. I took it to production.
The question we kept coming back to wasn't "can we build this?" — the models made building cheap. The question was: what do you build when building is cheap?
One Connection, Not Eight
When a design agency's data lives across 8 platforms, an agent needs 8 integrations, 8 auth flows, 8 different data schemas. Half these tools don't have APIs. The agent spends more time navigating tool boundaries than doing useful work.
When everything lives in one platform — projects, tasks, communication, files, invoices, design reviews, playbooks, CRM — an agent needs one connection. One permission model. One data layer. That's the architectural premise behind everything we built.
We put an AI assistant into the platform as a proof of concept — it has the same permissions as the user and can see everything. But the assistant is debugging disguised as a feature. The real play is making the platform accessible through a single API so users connect their own agents — through Claude Code, through whatever harness they're running — and operate on their entire agency.
A Design Agency Operating System
Every feature was designed around how agents would eventually need to interact with the data.
Sequence is everything
Every project has a unified timeline — a spinal cord for all tasks, communication, files, reviews, and milestones. Sequence matters tremendously in business operations. An agent generating a case study needs to know what happened before and after. An agent creating a playbook from a completed project needs to understand the order of operations. Without sequence, the agent can only summarize. With sequence, it can reason about process.
Visual assets need to be findable
Auto-tagging and AI-generated descriptions on every uploaded image. Multi-modal vector embeddings for visual asset retrieval. When a client asks for "that photo with the green label from the skincare shoot," someone needs to find it. Today that's a person scrolling through folders. The architecture makes it so an agent can retrieve it.
Feedback belongs on the design
Pin-based design reviews — feedback pinned to a pixel, threaded, with revisions tracked across rounds. Not in a WhatsApp message, not in an email chain. Structured feedback that designers and clients use today. Also the kind of structured data an agent can analyze, summarize, and flag unresolved items from.
SOPs that become task lists
Playbooks capture how the agency executes each service type — brand discovery, product photography, packaging design, website builds. When a new project starts, the playbook extracts into a task list. When a project completes, the system has enough context to generate a new playbook from what actually happened — capturing the real process, not just the planned one.
The accumulated context
Every conversation, every review round, every milestone, every file — structured and in one place. The system has enough context to generate case studies, blog posts, portfolio entries grounded in project specifics. The longer you use the platform, the more useful the data becomes. That's not something you get from a tool you built in a weekend.
The Bet
Most design agencies haven't changed their operations in any meaningful way beyond using a chatbot. More than half are still running on the same stack they had two years ago. Louie and I are further along the agent journey than most of our potential users are right now.
Our bet is that in a year, that changes. Agents will be on our phones. They'll be managing workflows, not just answering questions. When that happens, the agency running 8 disconnected tools will need to connect an agent to each one — or find a platform that already put everything in one place.
Software has never been cheaper to build. The moat isn't the code. It's understanding the problem deeply enough to design a system that gets more valuable over time. Louie has six years of that understanding. He's going to market with it now.